Thursday, October 9, 2025

4:00 pm-5:00 pm

Understanding Neural Networks and Their Applications in Physics

In this seminar, we will review the fundamentals of artificial neural networks, including key concepts such as artificial neurons, activation and loss functions, and backpropagation, which are essential building blocks of modern deep learning techniques. We will explore widely used architectures such as Multilayer Perceptrons, Convolutional Neural Networks, Graph Convolutional Neural Networks, Autoencoders, and Physics-Informed Neural Networks. Each architecture will be discussed in terms of its underlying principles and its suitability for different types of data and problems. Finally, we will present selected examples showcasing how deep learning has been applied to a variety of problems in physics.